Application method and system of personalized course recommendation system in education industry based on large model

By using a large-scale model-based personalized course recommendation system, the problem that traditional course recommendation systems cannot meet the unique needs of learners is solved, thus realizing personalized course recommendations and improving learning efficiency and resource utilization efficiency.

CN119048298APending Publication Date: 2024-11-29SHANDONG LANGCHAO YUNTOU INFORMATION TECH CO LTD
View PDF 0 Cites 6 Cited by

Patent Information

Application Number
CN202411118497.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-15
Publication Date
2024-11-29

AI Technical Summary

Technical Problem

Traditional course recommendation systems struggle to meet the unique needs of each learner, resulting in a lack of targeted recommendations that negatively impact learning efficiency and experience.

Method used

A personalized course recommendation system based on a large model is constructed through data collection and preprocessing, model building and fine-tuning, personalized recommendation algorithms, multimodal fusion processing, user interface and interaction, performance monitoring and optimization, and system security and privacy protection.

Benefits of technology

It enables customized course recommendations based on learners' needs and interests, improving learning efficiency and engagement, reducing unnecessary human and time investment, and promoting the fair distribution and efficiency improvement of educational resources.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119048298A_ABST
    Figure CN119048298A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of computer science and information, in particular to an application method and system of a personalized course recommendation system based on a large model in the education industry. Constructing and finely adjusting a large model; a personalized recommendation algorithm; carrying out multi-modal fusion processing; a user interface and interaction; monitoring and optimizing the performance; system security and privacy protection are realized; the application method and system of the personalized course recommendation system based on the large model in the education industry have the beneficial effects that through personalized recommendation, learners can find courses suitable for the current level and interest of the learners more quickly, invalid browsing and learning are avoided, and therefore the learning efficiency is improved; the recommendation system can provide related contents according to the needs and preferences of the learner, the interestingness and participation degree of learning are enhanced, and the learning satisfaction degree is improved; the system can dynamically adjust the recommendation according to the learning progress and feedback of the learner.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of computer science and information technology, and specifically to an application method and system of a large model-based personalized course recommendation system in the education industry. Background Art

[0002] With the rapid development of Internet technology and the widespread application of artificial intelligence technology, the education industry is undergoing a major transformation from the traditional model to digitalization and intelligence. The rise of emerging educational tools such as online education platforms and intelligent learning systems has greatly enriched the access to educational resources, while also placing higher demands on the personalization and precision of educational content. However, traditional course recommendation systems are mostly based on simple user behavior analysis and content association rules, which are difficult to fully meet the unique needs of each learner, resulting in the lack of pertinence in recommended content, which affects learning efficiency and experience.

[0003] In recent years, the emergence of deep learning, especially large models (such as Transformers), has provided new possibilities for personalized recommendations in the field of education. With its powerful representation learning ability, large models can capture complex user characteristics and content characteristics from massive data to achieve more accurate personalized matching. These models can learn rich language and knowledge structures through pre-training on large-scale data sets, and then after fine-tuning in specific fields, they can be applied to educational scenarios such as course recommendations and learning path planning.

[0004] Although big models have shown great potential in other industries, their application in personalized course recommendations in the education industry still faces challenges, including but not limited to: how to effectively integrate multi-source heterogeneous education data; how to design efficient and accurate recommendation algorithms to adapt to the special needs of educational scenarios; and how to build a system architecture that can both protect student privacy and continuously optimize recommendation effects.

[0005] Therefore, there is an urgent need for an innovative personalized course recommendation system that can fully utilize the advantages of large models, deeply understand learners' learning habits, ability levels, and interest preferences, and combine educational theories to provide customized course recommendations for each learner to promote their personalized learning and development. This is not only the key to improving the quality of education, but also an important way to achieve educational equity and efficiency. Summary of the invention

[0006] The purpose of the present invention is to provide an application method and system of a large model-based personalized course recommendation system in the education industry to solve the problems raised in the above-mentioned background technology.

[0007] To achieve the above object, the present invention provides the following technical solution: an application method of a personalized course recommendation system based on a large model in the education industry, the method comprising the following steps:

[0008] Data collection and preprocessing;

[0009] Large model building and fine-tuning;

[0010] Personalized recommendation algorithm;

[0011] Multimodal fusion processing;

[0012] User interface and interaction;

[0013] Performance monitoring and optimization;

[0014] System security and privacy protection.

[0015] Preferably, the specific operations of data collection and preprocessing include:

[0016] Through interfacing with learning management systems, online learning platforms, and social media data sources, we can comprehensively collect users’ learning behavior data, course information, and user preference data. Learning behavior data includes users’ browsing history, learning progress, and test scores; course information includes course name, course description, and course difficulty; user preferences include users’ interests, learning habits, and learning goals;

[0017] The collected data will be preprocessed, including data cleaning, missing value filling, outlier processing, and data standardization operations.

[0018] Preferably, the specific operations of large model construction and fine-tuning include:

[0019] The system selects BERT and RoBERTa pre-trained models as the basic models, and builds the models according to the characteristics of the education field, adding specific vocabulary in the education field, such as subject names and teaching methods; adjusts the structure of the model to adapt to the characteristics of the education field, such as the hierarchical structure of courses and the logical relationship of knowledge points; uses the corpus in the education field for model training to improve the performance of the model in the education field;

[0020] After the basic model is built, the system will use labeled education data to fine-tune the model. The purpose of fine-tuning is to enable the model to better understand and process education data and improve the accuracy of recommendations. During the fine-tuning process, methods such as cross entropy loss function and learning rate adjustment strategy are used to improve the generalization ability of the model.

[0021] Preferably, the specific operations of the personalized recommendation algorithm include:

[0022] The system adopts a hybrid recommendation strategy of collaborative filtering, content-based recommendation and deep learning model to provide users with personalized course recommendations. Collaborative filtering recommendation analyzes the similarities between users to find other users similar to the target user and recommend courses that these users like; content-based recommendation analyzes the attributes of the course, such as course description and course difficulty, to find courses that match the user's interests; deep learning model recommendation builds a semantic relationship between users and courses to find the courses that best meet the user's needs;

[0023] In order to improve the accuracy of recommendations, the system combines the above three recommendation algorithms. The fusion method performs weighted average of the recommendation results of different algorithms and uses a neural network structure to integrate the recommendation results of different algorithms.

[0024] Preferably, the specific operations of multimodal fusion processing include: the system will integrate text, image, and video multimodal information to improve the accuracy of recommendation, and the system will use multimodal learning methods, such as multi-channel convolutional neural networks and multimodal fusion networks, to process and fuse various information of users and courses; for text information, the system uses natural language processing technology, such as word segmentation and word vector representation, to extract semantic features of courses and users; for image information, the system uses convolutional neural networks to extract visual features of courses and users; for video information, the system uses video analysis technology to extract behavioral features of courses and users;

[0025] The specific operations of user interface and interaction include: the system designs user interface, provides personalized recommendations, and collects user feedback. The user interface is a web interface and a mobile application interface, which provides course recommendation, course search, and course evaluation functions; the system provides recommendation results to users through the user interface and collects user feedback. User feedback is user ratings, comments, and collection behaviors. The feedback will be used to optimize the recommendation algorithm and improve the accuracy of recommendations.

[0026] An application system of a large model-based personalized course recommendation system in the education industry, the system comprising:

[0027] The data collection layer integrates data from multiple sources such as the learning management system (LMS), online education platforms, social media, forums, and blogs to collect user learning behaviors, course information, and user preferences in real time;

[0028] The data preprocessing layer is responsible for data cleaning, missing value filling, outlier processing and data standardization to improve data quality;

[0029] The large model layer uses pre-trained models such as BERT or RoBERTa and fine-tunes them for the education field to enhance the model's understanding of educational content;

[0030] The feature extraction layer uses a large model to extract multi-dimensional features of users and courses, including semantic and structural information;

[0031] The recommendation algorithm layer combines collaborative filtering, content-based recommendation, and deep learning models to achieve personalized course recommendations;

[0032] The multimodal fusion layer integrates text, images, videos and other information, and uses a multimodal learning algorithm to improve the accuracy of recommendations;

[0033] User interaction layer, which provides personalized interface design and feedback collection mechanism to enhance user experience;

[0034] Performance optimization layer: implement real-time monitoring, A / B testing, and model iteration to continuously optimize recommendation results;

[0035] The security and privacy layer adopts data encryption, differential privacy technology and compliance management to ensure data security and privacy.

[0036] Preferably, the data preprocessing layer comprehensively collects users' learning behavior data, course information, and user preference data by interfacing with learning management systems, online learning platforms, and social media data sources. Learning behavior data includes users' browsing history, learning progress, and test scores; course information includes course name, course description, and course difficulty; user preferences include users' interests, learning habits, and learning goals; and the collected data will be preprocessed, including data cleaning, missing value filling, outlier processing, and data standardization operations.

[0037] Preferably, for the large model layer, the system selects BERT and RoBERTa pre-trained models as the basic models, and constructs the model according to the characteristics of the education field, adding specific vocabulary in the education field, such as subject names and teaching methods; adjusting the structure of the model to adapt to the characteristics of the education field, such as the hierarchy of courses, the logical relationship between knowledge points, etc.; using the corpus in the education field for model training to improve the performance of the model in the education field; after the basic model is built, the system will use the labeled education field data to fine-tune the model. The purpose of fine-tuning is to enable the model to better understand and process education field data and improve the accuracy of recommendations. During the fine-tuning process, strategies such as cross entropy loss function and learning rate adjustment are used to improve the generalization ability of the model.

[0038] Preferably, at the recommendation algorithm layer, the system adopts a hybrid recommendation strategy of collaborative filtering, content-based recommendation and deep learning model to provide users with personalized course recommendations. Collaborative filtering recommendation analyzes the similarities between users, finds other users similar to the target user, and recommends courses that these users like; content-based recommendation analyzes the attributes of the course, such as course description and course difficulty, to find courses that match the user's interests; deep learning model recommendation builds a semantic relationship between the user and the course to find the course that best meets the user's needs; in order to improve the accuracy of the recommendation, the system integrates the above three recommendation algorithms. The fusion method performs a weighted average of the recommendation results of different algorithms and uses a neural network structure to integrate the recommendation results of different algorithms.

[0039] Preferably, in the multimodal fusion layer, the system will integrate text, image, and video multimodal information to improve the accuracy of recommendations. The system will use multimodal learning methods, such as multi-channel convolutional neural networks and multimodal fusion networks, to process and fuse various information about users and courses. For text information, the system uses natural language processing techniques, such as word segmentation and word vector representation, to extract semantic features of courses and users. For image information, the system uses convolutional neural networks to extract visual features of courses and users. For video information, the system uses video analysis technology to extract behavioral features of courses and users.

[0040] At the user interaction layer, the system designs a user interface, provides personalized recommendations, and collects user feedback. The user interface is a web interface and a mobile application interface, which provides course recommendation, course search, and course evaluation functions. The system provides recommendation results to users through the user interface and collects user feedback. User feedback is user ratings, comments, and collection behaviors. The feedback will be used to optimize the recommendation algorithm and improve the accuracy of recommendations.

[0041] Compared with the prior art, the present invention has the following beneficial effects:

[0042] The invention proposes an application method and system of a large-model-based personalized course recommendation system in the education industry. Through personalized recommendations, learners can find courses suitable for their current level and interests more quickly, avoid ineffective browsing and learning, and thus improve learning efficiency; the recommendation system can provide relevant content according to the needs and preferences of learners, enhance the interest and participation of learning, and improve learning satisfaction; the system can dynamically adjust recommendations according to the learners' learning progress and feedback, help build a reasonable knowledge structure, and promote all-round development; through accurate recommendations, unnecessary manpower and time investment can be reduced, and the utilization efficiency of educational resources can be improved; personalized recommendations enable high-quality educational resources to be distributed more fairly, which helps to narrow the learning gap between different students. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 is a flow chart of the method of the present invention;

[0044] Figure 2 This is a system block diagram of the present invention. DETAILED DESCRIPTION

[0045] In order to make the purpose and technical solution of the present invention clearly and completely described, and the advantages more clearly understood, the embodiments of the present invention are further described in detail with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are part of the embodiments of the present invention, rather than all of the embodiments, and are only used to explain the embodiments of the present invention, and are not used to limit the embodiments of the present invention. All other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0046] For example, see Figure 1 The present invention provides a technical solution: an application method of a large model-based personalized course recommendation system in the education industry, the method comprising the following steps:

[0047] 1. System architecture

[0048] (1) Data Collection Layer

[0049] - Diversified data sources: Integrate data from multiple channels such as LMS, online education platforms, social media, forums, blogs, etc., including learning activities, discussions, comments, ratings, etc.

[0050] -Real-time update: Use real-time stream processing technology to continuously capture new user behaviors and course information.

[0051] (2) Data preprocessing layer

[0052] -Data cleaning: remove noisy data, eliminate duplicates, and correct errors.

[0053] -Missing value handling: using statistical methods, interpolation or model-based filling strategies.

[0054] -Outlier detection and processing: Use statistical and machine learning methods to identify and process outliers.

[0055] -Data standardization and normalization: unify the scales of different features to facilitate model training.

[0056] (3) Large model layer

[0057] -Model selection and customization: Use pre-trained models such as BERT and RoBERTa, and customize them for specific vocabulary and context in the education field.

[0058] - Fine-tuning strategy: Use the education field corpus to conduct targeted training of the model to improve the model's understanding of educational content.

[0059] (4) Feature extraction layer

[0060] -Semantic understanding: Extracting user interest patterns and abstract features of courses through large models.

[0061] -Multi-granularity features: Extract multi-level features such as course structure information, teacher reputation, user comment sentiment, etc.

[0062] (5) Recommendation algorithm layer

[0063] -Collaborative filtering: Based on user historical behaviors, discover similarities between users and recommend courses that similar users like.

[0064] -Content-based recommendations: Match user interests with course content features to achieve personalized recommendations based on content relevance.

[0065] -Deep learning models: such as deep neural networks (DNNs) and convolutional neural networks (CNNs), combined with multi-layer representation learning, improve the accuracy of recommendations.

[0066] (6) Multimodal Fusion Layer

[0067] -Multi-source data integration: Integrate multimodal data such as text, images, and videos into the recommendation system.

[0068] - Modal feature fusion: Use multimodal learning algorithms, such as multimodal attention mechanism, to fuse features of different modalities.

[0069] (7) User Interaction Layer

[0070] -Personalized interface design: Customize the interface based on user preferences and behavior patterns, and provide dynamic layout and content recommendations.

[0071] -User feedback system: collect user feedback on recommendations through ratings, comments, likes, etc.

[0072] (8) Performance Optimization Layer

[0073] -Real-time monitoring: Real-time tracking of system performance, including key indicators such as recommendation speed and accuracy.

[0074] -A / B testing: Conduct experiments regularly to compare the effectiveness of different recommendation strategies and select the best practices.

[0075] -Model iteration: Dynamically update model parameters and recommendation strategies based on user feedback and performance data.

[0076] (9) Security and Privacy Layer

[0077] -Data encryption: Use SSL / TLS and other protocols to encrypt data transmission to ensure the security of data during transmission.

[0078] -Privacy protection: Apply differential privacy technology to protect the privacy of user data.

[0079] - Compliance: Follow GDPR, Cybersecurity Law and other regulations to ensure the legality of data processing and storage.

[0080] 2. Data Collection and Preprocessing

[0081] (1) Data collection layer: The system interfaces with data sources such as learning management systems (LMS), online learning platforms, and social media to comprehensively collect user learning behavior data, course information, user preferences, and other data. Learning behavior data includes user browsing history, learning progress, and test scores; course information includes course name, course description, and course difficulty; and user preferences include user interests, learning habits, and learning goals.

[0082] (2) Data preprocessing layer: In order to improve the quality of data, the system will preprocess the collected data. Preprocessing includes data cleaning, missing value filling, outlier processing, data standardization and other operations. Data cleaning is to remove invalid data and noise to improve the authenticity of data; missing value filling is to deal with missing values ​​in the data, and commonly used methods include mean filling and median filling; outlier processing is to identify and process outliers in the data to improve the accuracy of the data; data standardization is to convert data into numerical values ​​with a unified scale for subsequent processing.

[0083] 3. Large model construction and fine-tuning

[0084] (1) Large model construction: The system selects pre-trained models such as BERT and RoBERTa as the basic model, and constructs the model according to the characteristics of the education field. Specifically, it can be constructed from the following aspects: adding specific vocabulary in the education field, such as subject names, teaching methods, etc.; adjusting the structure of the model to adapt to the characteristics of the education field, such as the hierarchy of courses, the logical relationship of knowledge points, etc.; using the corpus in the education field for model training to improve the performance of the model in the education field.

[0085] (2) Fine-tuning of large models: After the basic model is built, the system will use the labeled education data to fine-tune the model. The purpose of fine-tuning is to enable the model to better understand and process education data and improve the accuracy of recommendations. During the fine-tuning process, strategies such as cross entropy loss function and learning rate adjustment can be used to improve the generalization ability of the model.

[0086] 4. Personalized recommendation algorithm

[0087] (1) Personalized recommendation algorithm: The system adopts a hybrid recommendation strategy of collaborative filtering, content-based recommendation and deep learning model to provide users with personalized course recommendations. Collaborative filtering recommendation analyzes the similarity between users to find other users similar to the target user and recommend courses that these users like; content-based recommendation analyzes the attributes of the course, such as course description, course difficulty, etc., to find courses that match the user's interests; deep learning model recommendation builds a semantic relationship between users and courses to find the courses that best meet the user's needs.

[0088] (2) Algorithm fusion: In order to improve the accuracy of recommendations, the system combines the above three recommendation algorithms. The fusion method can be to take a weighted average of the recommendation results of different algorithms, or to use a neural network structure to integrate the recommendation results of different algorithms.

[0089] 5. Multimodal fusion processing

[0090] (1) Multimodal fusion layer: The system will integrate multimodal information such as text, images, and videos to improve the accuracy of recommendations. Specifically, the system will use multimodal learning methods, such as multi-channel convolutional neural networks and multimodal fusion networks, to process and fuse various information about users and courses.

[0091] (2) Modal information processing: For text information, the system can use natural language processing techniques, such as word segmentation and word vector representation, to extract the semantic features of courses and users. For image information, the system can use convolutional neural networks to extract the visual features of courses and users. For video information, the system can use video analysis technology to extract the behavioral features of courses and users.

[0092] 6. User Interface and Interaction

[0093] (1) User interaction layer: The system designs a user interface, provides personalized recommendations, and collects user feedback. The user interface can be a web interface, a mobile application interface, etc., and provides functions such as course recommendation, course search, and course evaluation.

[0094] (2) User feedback collection: The system provides recommendation results to users through the user interface and collects user feedback. User feedback can be user ratings, comments, favorites, etc. These feedbacks will be used to optimize the recommendation algorithm and improve the accuracy of recommendations.

[0095] 7.Performance monitoring and optimization

[0096] (1) Performance optimization layer: The system monitors system performance in real time and uses A / B testing and model iteration to optimize recommendation results. Through A / B testing, the system can compare the results of different recommendation algorithms and select the optimal algorithm; through model iteration, the system can continuously optimize model parameters and improve the accuracy of recommendations.

[0097] (2) Resource Scheduling: In order to improve system performance, the system will perform reasonable resource scheduling based on user needs and system resource conditions. For example, during peak user hours, server resources can be increased to ensure stable system operation; during periods of low resource utilization, server resources can be appropriately reduced to save costs.

[0098] 8. System security and privacy protection

[0099] (1) Security and privacy layer: The system implements data encryption and privacy protection measures to ensure the security and privacy of user data. Data encryption can use SSL encryption technology to protect the security of data during transmission; privacy protection measures can include access control, data desensitization, user privacy settings, etc. to ensure the security of user data during use.

[0100] (2) Compliance Check: The system will also conduct regular compliance checks to ensure that data processing and recommendation services comply with relevant laws and regulations. This includes but is not limited to data protection regulations (such as GDPR), education industry standards, and user agreements.

[0101] (3) Security Audit: The system should implement security audits and record the history of data access and modification so that the specific operations can be traced when a security incident occurs. In addition, the system should regularly test and evaluate security measures to ensure their effectiveness.

[0102] (4) User control: In order to improve users’ trust in the recommendation system, the system should provide user control functions to allow users to manage their own privacy settings, such as choosing whether to participate in recommendations, viewing and modifying personal data, and controlling data sharing.

[0103] Embodiment 2, based on embodiment 1, proposes an application system of a personalized course recommendation system based on a large model in the education industry, the system comprising:

[0104] The data collection layer integrates data from multiple sources such as the learning management system (LMS), online education platforms, social media, forums, and blogs to collect user learning behaviors, course information, and user preferences in real time;

[0105] The data preprocessing layer is responsible for data cleaning, missing value filling, outlier processing and data standardization to improve data quality; by interfacing with learning management systems, online learning platforms, and social media data sources, it comprehensively collects users' learning behavior data, course information, and user preference data. Learning behavior data includes users' browsing history, learning progress, and test scores; course information includes course name, course description, and course difficulty; user preferences include users' interests, learning habits, and learning goals; the collected data will be preprocessed, including data cleaning, missing value filling, outlier processing, and data standardization operations.

[0106] At the large model layer, pre-trained models such as BERT or RoBERTa are used, and fine-tuned for the education field to enhance the model's understanding of educational content. The system selects BERT and RoBERTa pre-trained models as basic models, and builds the model according to the characteristics of the education field, adding specific vocabulary in the education field, such as subject names and teaching methods. The structure of the model is adjusted to adapt to the characteristics of the education field, such as the hierarchy of courses and the logical relationship between knowledge points. The model is trained using a corpus in the education field to improve the performance of the model in the education field. After the basic model is built, the system will use labeled education field data to fine-tune the model. The purpose of fine-tuning is to enable the model to better understand and process education field data and improve the accuracy of recommendations. During the fine-tuning process, methods such as cross-entropy loss function and learning rate adjustment strategy are used to improve the generalization ability of the model.

[0107] The feature extraction layer uses a large model to extract multi-dimensional features of users and courses, including semantic and structural information;

[0108] The recommendation algorithm layer combines collaborative filtering, content-based recommendation and deep learning models to achieve personalized course recommendations. The system adopts a hybrid recommendation strategy of collaborative filtering, content-based recommendation and deep learning models to provide users with personalized course recommendations. Collaborative filtering recommendation analyzes the similarities between users to find other users similar to the target user and recommend courses that these users like. Content-based recommendation analyzes the attributes of the course, such as course description and course difficulty, to find courses that match the user's interests. Deep learning model recommendation builds a semantic relationship between users and courses to find the courses that best meet the user's needs. In order to improve the accuracy of the recommendation, the system integrates the above three recommendation algorithms. The fusion method takes a weighted average of the recommendation results of different algorithms and uses a neural network structure to integrate the recommendation results of different algorithms.

[0109] The multimodal fusion layer integrates text, image, video and other information, and uses a multimodal learning algorithm to improve the accuracy of recommendations. The system will integrate text, image, and video multimodal information to improve the accuracy of recommendations. The system will use multimodal learning methods, such as multi-channel convolutional neural networks and multimodal fusion networks, to process and fuse various information about users and courses. For text information, the system uses natural language processing technology, such as word segmentation and word vector representation, to extract semantic features of courses and users. For image information, the system uses convolutional neural networks to extract visual features of courses and users. For video information, the system uses video analysis technology to extract behavioral features of courses and users.

[0110] The user interaction layer provides personalized interface design and feedback collection mechanism to enhance user experience; the system designs the user interface, provides personalized recommendations, and collects user feedback. The user interface is a web interface and a mobile application interface, which provides course recommendation, course search, and course evaluation functions; the system provides recommendation results to users through the user interface and collects user feedback. User feedback is user ratings, comments, and collection behaviors. The feedback will be used to optimize the recommendation algorithm and improve the accuracy of recommendations.

[0111] The performance optimization layer implements real-time monitoring, A / B testing, and model iteration to continuously optimize recommendation results.

[0112] The security and privacy layer adopts data encryption, differential privacy technology and compliance management to ensure data security and privacy.

[0113] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An application method of a large model-based personalized course recommendation system in the education industry, characterized by: The method comprises the following steps: Data collection and preprocessing; Large model building and fine-tuning; Personalized recommendation algorithm; Multimodal fusion processing; User interface and interaction; Performance monitoring and optimization; System security and privacy protection.

2. The application method of the large model-based personalized course recommendation system in the education industry according to claim 1 is characterized by: The specific operations of data collection and preprocessing include: Through interfacing with learning management systems, online learning platforms, and social media data sources, we can comprehensively collect users’ learning behavior data, course information, and user preference data. Learning behavior data includes users’ browsing history, learning progress, and test scores; course information includes course name, course description, and course difficulty; user preferences include users’ interests, learning habits, and learning goals; The collected data will be preprocessed, including data cleaning, missing value filling, outlier processing, and data standardization operations.

3. The application method of the large model-based personalized course recommendation system in the education industry according to claim 1 is characterized by: The specific operations of large model construction and fine-tuning include: The system selects BERT and RoBERTa pre-trained models as the basic models, and builds the models according to the characteristics of the education field, adding specific vocabulary in the education field, such as subject names and teaching methods; adjusts the structure of the model to adapt to the characteristics of the education field, such as the hierarchical structure of courses and the logical relationship of knowledge points; uses the corpus in the education field for model training to improve the performance of the model in the education field; After the basic model is built, the system will use labeled education data to fine-tune the model. The purpose of fine-tuning is to enable the model to better understand and process education data and improve the accuracy of recommendations. During the fine-tuning process, methods such as cross entropy loss function and learning rate adjustment strategy are used to improve the generalization ability of the model.

4. The application method of the large model-based personalized course recommendation system in the education industry according to claim 1 is characterized by: The specific operations of the personalized recommendation algorithm include: The system adopts a hybrid recommendation strategy of collaborative filtering, content-based recommendation and deep learning model to provide users with personalized course recommendations. Collaborative filtering recommendation analyzes the similarities between users to find other users similar to the target user and recommend courses that these users like; content-based recommendation analyzes the attributes of the course, such as course description and course difficulty, to find courses that match the user's interests; deep learning model recommendation builds a semantic relationship between users and courses to find the courses that best meet the user's needs; In order to improve the accuracy of recommendations, the system combines the above three recommendation algorithms. The fusion method performs weighted average of the recommendation results of different algorithms and uses a neural network structure to integrate the recommendation results of different algorithms.

5. The application method of the large model-based personalized course recommendation system in the education industry according to claim 1 is characterized by: The specific operations of multimodal fusion processing include: the system will integrate text, image, and video multimodal information to improve the accuracy of recommendations. The system will use multimodal learning methods, such as multi-channel convolutional neural networks and multimodal fusion networks, to process and fuse various information about users and courses. For text information, the system uses natural language processing techniques, such as word segmentation and word vector representation, to extract the semantic features of courses and users. For image information, the system uses convolutional neural networks to extract the visual features of courses and users. For video information, the system uses video analysis technology to extract the behavioral features of courses and users. The specific operations of user interface and interaction include: the system designs user interface, provides personalized recommendations, and collects user feedback. The user interface is a web interface and a mobile application interface, which provides course recommendation, course search, and course evaluation functions; the system provides recommendation results to users through the user interface and collects user feedback. User feedback is user ratings, comments, and collection behaviors. The feedback will be used to optimize the recommendation algorithm and improve the accuracy of recommendations.

6. An application system of a personalized course recommendation system based on a big model in the education industry according to the application method of the personalized course recommendation system based on a big model in the education industry according to any one of claims 1 to 5, characterized in that: The system comprises: The data collection layer integrates data from multiple sources such as the learning management system (LMS), online education platforms, social media, forums, and blogs to collect user learning behaviors, course information, and user preferences in real time; The data preprocessing layer is responsible for data cleaning, missing value filling, outlier processing and data standardization to improve data quality; The large model layer uses pre-trained models such as BERT or RoBERTa and fine-tunes them for the education field to enhance the model's understanding of educational content; The feature extraction layer uses a large model to extract multi-dimensional features of users and courses, including semantic and structural information; The recommendation algorithm layer combines collaborative filtering, content-based recommendation, and deep learning models to achieve personalized course recommendations; The multimodal fusion layer integrates text, images, videos and other information, and uses a multimodal learning algorithm to improve the accuracy of recommendations; User interaction layer, which provides personalized interface design and feedback collection mechanism to enhance user experience; Performance optimization layer: implement real-time monitoring, A / B testing, and model iteration to continuously optimize recommendation results; The security and privacy layer adopts data encryption, differential privacy technology and compliance management to ensure data security and privacy.

7. The application system of the large model-based personalized course recommendation system in the education industry according to claim 6, characterized in that: The data preprocessing layer comprehensively collects users’ learning behavior data, course information, and user preference data by interfacing with learning management systems, online learning platforms, and social media data sources. Learning behavior data includes users’ browsing history, learning progress, and test scores. Course information includes course name, course description, and course difficulty; user preferences include user interests, learning habits, and learning goals; the collected data will be preprocessed, including data cleaning, missing value filling, outlier processing, and data standardization operations.

8. The application system of the large model-based personalized course recommendation system in the education industry according to claim 6, characterized in that: At the large model level, the system selects BERT and RoBERTa pre-trained models as the basic models, and builds the models based on the characteristics of the education field, adding specific vocabulary in the education field, such as subject names and teaching methods; and adjusting the structure of the model to adapt to the characteristics of the education field, such as the hierarchical structure of courses and the logical relationship between knowledge points. Use the corpus in the education field to train the model and improve the performance of the model in the education field; After the basic model is built, the system will use labeled education data to fine-tune the model. The purpose of fine-tuning is to enable the model to better understand and process education data and improve the accuracy of recommendations. During the fine-tuning process, methods such as cross entropy loss function and learning rate adjustment strategy are used to improve the generalization ability of the model.

9. The application system of the large model-based personalized course recommendation system in the education industry according to claim 6, characterized in that: At the recommendation algorithm layer, the system adopts a hybrid recommendation strategy of collaborative filtering, content-based recommendation and deep learning model to provide users with personalized course recommendations. Collaborative filtering recommendation analyzes the similarities between users, finds other users similar to the target user, and recommends courses that these users like; content-based recommendation analyzes the attributes of the course, such as course description and course difficulty, to find courses that match the user's interests; deep learning model recommendation builds a semantic relationship between users and courses to find the courses that best meet the user's needs; in order to improve the accuracy of the recommendation, the system integrates the above three recommendation algorithms. The fusion method takes a weighted average of the recommendation results of different algorithms and uses a neural network structure to integrate the recommendation results of different algorithms.

10. The application system of the large model-based personalized course recommendation system in the education industry according to claim 6, characterized in that: In the multimodal fusion layer, the system will integrate text, image, and video multimodal information to improve the accuracy of recommendations. The system will use multimodal learning methods, such as multi-channel convolutional neural networks and multimodal fusion networks, to process and fuse various information about users and courses. For text information, the system uses natural language processing techniques, such as word segmentation and word vector representation, to extract semantic features of courses and users. For image information, the system uses convolutional neural networks to extract visual features of courses and users. For video information, the system uses video analysis technology to extract course and user behavior characteristics; At the user interaction layer, the system designs a user interface, provides personalized recommendations, and collects user feedback. The user interface is a web interface and a mobile application interface, which provides course recommendation, course search, and course evaluation functions. The system provides recommendation results to users through the user interface and collects user feedback. User feedback is user ratings, comments, and collection behaviors. The feedback will be used to optimize the recommendation algorithm and improve the accuracy of recommendations.

Citation Information

Cited By

  • Teaching courseware intelligent recommendation method

    CN119311954A

  • Generative man-machine fusion student portrait analysis method and system, terminal and medium

    CN119762302A

  • Clustering algorithm-based education platform course optimization recommendation system and method

    CN120144877A

  • Course recommendation method and system based on large language model decoupling modeling

    CN120197727A

  • A Course Recommendation Method and System Based on Decoupling Modeling of Large Language Models

    CN120197727B